数据堂—87,871张人脸106关键点标注数据（复杂场景）

87,871张人脸106关键点标注数据（复杂场景）包含黄种人、黑人、白人和印度人。为增加数据难度，数据多样性包括脸部不同姿态，不同表情，不同光照，不同场景。87,871张人脸106关键点标注数据可用于人脸关键点定位、人脸行为识别等任务。

87,871 facial images with 106-point facial landmark annotations (in complex scenes) cover East Asians, Black people, Caucasians and Indians. To enhance the dataset's difficulty, the data features diverse variations including different facial poses, expressions, lighting conditions and scenes. This annotated facial landmark dataset can be applied to tasks such as facial landmark localization and facial behavior recognition.

搜集汇总

背景与挑战

背景概述

该数据集包含87,871张人脸图像，每张图像均标注了106个关键点和9种属性，覆盖多种人种、性别、年龄、姿态、表情、光照和复杂场景，适用于人脸关键点定位和行为识别任务。数据准确率超过97%，格式包括.jpg和.json，并具有多样化的采集环境。